TCGA survival data classification

Hello everyone,

I am working on one of the Ovarian cancer TCGA clinical data . I would like to classify patients into two classes based on the survival. I took the column days to death (short : less than three years survival (3653) and long : > 3653 : greater than three year survival and classified the data.

It is also mentioned to use of days to death or days to last follow up for survival analysis (http://www.omnesres.com/tools/tcga/.

I am following one the publication where number of short and long survival are different from my observed count. https://academic.oup.com/jamia/article/22/1/109/832766.

Other study : (147 ST /193 LT )

In my case I have : ( ST 178/166 LT)

In my case, I just took column days to death and classified data into two classes. There are other columns are available such as "yearstobirth" "vitalstatus" "daystodeath" "daystolastfollowup"

I will appreciate all suggestions in case, if my approach need to be adapted.

Thanks

tcga survival

• 4,719 views • link

updated 6.9 years ago by Kevin Blighe

• written 6.9 years ago by Omics data mining

1 answer

Firstly, I would not worry too much about your patient numbers not matching. The TCGA data is constantly evolving and you will spend days or weeks trying to trace back the discrepancies. Just take the data that you currently have but date-stamp it and provide clear details as to how you obtained it.

With regard to overall survival, you can calculate it as date from pathologic diagnosis to date of death or last follow-up. So:

OS <- as.integer(
  ifelse( is.na(clindata$days_to_death),
    clindata$days_to_last_followup,
    clindata$days_to_death))

For PFS, take a look at the columns:

days_to_new_tumor_event_after_initial_treatment
days_to_additional_surgery_metastatic_procedure

Kevin

• 0 views • link

written 6.9 years ago by Kevin Blighe

Thanks alot Kevin. It helped alot.

• 0 views • link

written 6.9 years ago by Omics data mining